LlamaIndex publishes a 37-page technical report on OCRia-1, its document-parsing model
On Oct 22 LlamaIndex released a 37-page technical report on OCRia-1, a 4B-parameter document-parsing VLM. It reports a mean ParseBench score of 0.87 across four dimensions, against 0.75 for the next-best system. On one B200 GPU it decodes about 4,700 tokens/s on a single request; its per-request decoding speed is 3.4x and 2.0x that of the next-fastest rival at one and eight concurrent requests.
Why it matters
LlamaIndex published a 37-page OCRia-1 technical report with multi-benchmark comparisons against major open-weight OCR models and details of its data and training pipeline.
Sources2
LlamaIndexMediaOCRia-1 Technical ReportView original ·
- @jerryjliu0XWe wrote a 37-page technical report on OCRia-1, our SOTA model for document parsing. ✅ It contains comprehensive evals against most other major open-weight OCR models (e.g. Infinity-Parser, LightOnOCR 3, Chandra 2, dots.mocr) on multiple OCR benchmarks: ParseBench, olmOCR-bench, PulseBench-Tab, fr-bench-pdf2md. ✅ It contains detailed visibility into our data and training pipeline, to give a glimpse into how we’re able to curate the right document datasets to push the evaluation metrics within dimensions that matter for doc parsing. ✅ It is blazing fast ⚡️ decoding 2x-4x+ faster than the latest open-weight models. ✅ If you’re building your own vision models, there are 24 pages worth of appendix details (!!) that showcase all of our experimental findings from training to evals! Check it out: https://www.llamaindex.ai/ocria/ocria-1-technical-report.pdf Other links are in our blog! https://www.llamaindex.ai/blog/parse-ocria-1View original ·
Written by dalao.ai: the title, summary and why it matters are our own words. The original reports are the record; their publishers keep authorship.